Intel has unveiled its latest Xeon CPU roadmap reaching through to 2024, revealing a distinct focus on power efficiency.
It follows a general trend of chip manufacturers focusing more heavily on performance per-watt, a strategy that Intel said would underpin continued growth in the data centre market, as well as cloud, network, and edge computing.
The new “ultra-efficient” ‘Sierra Forest’ processor family, due for release in 2024 on the Intel 3 process technology, will upgrade key products to more advanced process nodes and introduce a wide-ranging architecture strategy for the data centre, it said.
Intel will launch a series of processors before Sierra Forest reaches the market, starting with the P-core Sapphire Rapids in March 2022, built on the Intel 7 process technology.
Sapphire Rapids will deliver up to a 30x performance increase in artificial intelligence compute, Intel said, and will represent the most feature-rich Xeon chip the company has released to date.
Shipments for the new chip will begin next month and support technologies such as DDR5 memory, PCIe Gen5 for double the bandwidth over PCIe Gen4, and the Compute Express Link standard for high-speed chip-to-device connections.
Emerald Rapids will mark the next stage of the Xeon roadmap and will start to ship in 2023, acting as a refresh to Sapphire Rapids, boosting performance while improving the platform’s memory and security capabilities.
Along with Sierra Forest, Granite Rapids will succeed Sapphire Rapids and will also be upgraded in 2024 from the Intel 4 to the Intel 3 process.
“Today we're giving the world a clear view of our industry-leading Xeon roadmap, which will fuel our growth and leadership through 2024 and beyond,” said Sandra Rivera, executive vice president and general manager of the Datacenter and AI Group at Intel.
“This diversified product portfolio has been developed in tight partnership with our customers, based on their diverse needs, aligned to their timelines and designed to fuel even more innovations with their developer ecosystems.”
Arc desktop GPUs timeline revealed
Modernise and thrive with Device-as-a-Service
Improving end-user experience through modernisation
Intel has also provided updated timelines for an array of other products across it business units, including the long-awaited Arc GPUs providing discrete graphics to laptops and PCs.
The company said it expects to ship more than 4 million discrete GPUs in 2022, starting with Arc GPUs for laptops which are coming in Q1 2022, while the desktop version will be launching slightly later at some point on Q2. Workstation chips will coming even later in Q3.
In addition, Intel announced it has already begun work developing the third generation of Arc GPUs, currently dubbed ‘Celestial’. The products will aim to address the “ultra-enthusiast” corner of the market, indicating a potential shift to rival Nvidia and AMD’s most powerful graphics units.
Elsewhere, Intel teased what it calls ‘Project Endgame’ - a cloud-based GPU service promising an “always-accessible low-latency computing experience”. Very few details about this project are currently known but we know it’s due to launch later this year.
